Search papers, labs, and topics across Lattice.
This paper investigates overlap bias in LLM judges when evaluating text summaries, revealing a tendency to favor LLM-generated summaries over human-written summaries as the similarity (ROUGE, BLEU) between the judged summaries decreases. The study tested 9 LLMs (1B-12B parameters, including Gemma 3 and LLaMA 3 variants) and found that this bias persists across most models, irrespective of their inherent positional biases. The results highlight the limitations of using LLMs as judges in summarization tasks based solely on overlap-based comparisons.
LLM judges exhibit a surprising "blindness" to human-written summaries, increasingly preferring machine-generated content as the similarity to human references decreases, challenging their reliability in summarization tasks.
Large language model (LLM) judges have often been used alongside traditional, algorithm-based metrics for tasks like summarization because they better capture semantic information, are better at reasoning, and are more robust to paraphrasing. However, LLM judges show biases for length and order among others, and are vulnerable to various adversarial input prompts. While recent studies have looked into these biases, few have analyzed them at a more granular level in relation to a well-defined overlap metric. In this work we provide an LLM judge bias analysis as a function of overlap with human-written responses in the domain of summarization. We test 9 recent LLMs with parameter counts ranging from 1 billion to 12 billion, including variants of Gemma 3 and LLaMA 3. We find that LLM judges increasingly prefer summaries generated by other LLMs over those written by humans as the similarities (as measured by ROUGE and BLEU) between the judged summaries decrease, and this pattern extends to all but one model tested, and exists regardless of the models'own position biases. Additionally, we find that models struggle to judge even summaries with limited overlaps, suggesting that LLM-as-a-judge in the summary domain should rely on techniques beyond a simple comparison.
